Adam FordPortfolio / Agentic Systems
6 systems in production
Built withv0LinkedIn
Professional Services · Built agentically

I connect business problems to working systems.

Six tools that replaced hours of manual toil across proposals, pipeline, resourcing, and storytelling. Each one was designed, built, and shipped through agentic coding, using LLMs, MCP, and real-time data capabilities that didn't exist a few years ago. This site was too. I built it with Vercel v0.

  • 6
    Systems shipped
  • 3
    Agent skills
  • Real-time
    BigQuery data
  • Hrs → Min
    Time to output

One builder. Six systems. A shared platform.

Every system sits on the same set of primitives. Hover or focus a node to trace what it's wired to, and select a system to jump to its architecture.

Layer 01 / SystemsBusiness-facing
Layer 02 / Platform primitivesShared infrastructure

The systems, problem to impact.

Each one starts with a real business problem, maps to an architecture, and ends in a measurable change in how people work.

01

EM Agent Skills Library

A portable library of agent skills that lets Engagement Managers do hours of pre-sales and contracting work in minutes, in any harness they choose.

Agent skills · Multi-harness
Problem

Proposals and SOWs took hours each to draft by hand, and limited prospecting time was spread across every merchant in a vertical.

What I built

Three skills: Proposal, Proposal → SOW, and Hunt. They source context across MCP tools and the warehouse, then generate grounded, on-format documents and ranked targets.

Impact

Hours of drafting per proposal and SOW became minutes, and the output was more comprehensive than before. EMs now spend their time on taste and judgment.

Before: Hours per proposal & SOW
After: Minutes, more comprehensive
Architecture / Skills LibraryInputs → Core → Outputs
Sources
MCP
Google Drive
MCP
Slack
MCP
GitHub
API
Data warehouse
API
Web search
Skill runtimePi + LLMs
Proposal skill
Target merchant → full draft with aligned services & pricing
Proposal → SOW
Negotiated proposal → contract-ready SOW schedule
Hunt skill
Onboards to an EM's verticals and surfaces the pre-sales merchants with the highest need for Services
Context corpora
Services catalog · past proposals · prior SOWs · legal guidelines · output design
Outputs
Doc
Drafted proposal
Contract
SOW schedule
Signal
Ranked target merchants
  • LLMs
  • Pi Harness
  • MCP
  • BigQuery
  • GitHub
  • Pi harness
  • Anthropic
  • OpenAI
  • Data warehouse API
02

Revenue & Pipeline Dashboard

The single source of truth for business performance. It's a retro, 90s-style terminal dashboard over live warehouse data, with one toggle between Closed Won and Open.

Real-time BI · AI narrative
Problem

Performance lived in one-off reports and spreadsheets, and deal categorization didn't map cleanly to the services catalog.

What I built

A live dashboard with KPI blocks, service-line splits, revenue pacing charts, and a deal table linked to CRM and contracts. It adds agentic deal categorization with manual overrides and an AI ‘Storyize’ review.

Impact

No more one-off reports or spreadsheet wrangling. It became the source of truth for telling the story of closed business and open pipeline.

Before: One-off reports & spreadsheets
After: Live single source of truth
Architecture / Pipeline DashboardInputs → Core → Outputs
Sources
SQL
BigQuery (real-time)
Link
CRM opportunities
Link
Contract system
Dashboard engineCached SQL
KPI blocks
Revenue vs target · deals · conversion · avg size
Service lines
Technical · Growth · Payment Migrations
Time-series charts
By year, region, segment, craft + monthly pacing
Agentic categorization
LLM maps deal names to catalog services, with overrides
Storyize
On-demand AI quarterly summary with prior-year comparisons
Outputs
Review
On-demand business review
Attribution
Service-line performance
Table
Deal rows → CRM & contract
  • Quick
  • BigQuery
  • LLMs
  • GitHub
  • Caching
  • LLM summaries
03

Book of Business

An always-on executive view of the Engagement Management book. You can see the whole team at a glance, then drill into any EM's closed performance and open pipeline.

Always-on executive site
Problem

Every leadership meeting needed new slides, updated by hand to pull out key opportunities and next actions.

What I built

A live site with left navigation across EMs and Closed / Open / Asks tabs. EMs log asks for leadership with owners and progress, stored so other tools can query them.

Impact

Zero deck prep for every meeting it powers. The leadership asks that unlock the book are surfaced and tracked where leaders already look.

Before: Manual slides every meeting
After: Always-on, zero prep
Architecture / Book of BusinessInputs → Core → Outputs
Sources
SQL
BigQuery (real-time)
Input
EM-submitted asks
Executive presentation layerLive
EM navigator
Left nav: team summary → individual book
Closed
Won by quarter, top deals by revenue in focus
Open
Pipeline by quarter with next actions
Asks
Owners, status, and progress tracking
Asks persisted as JSON in quick.db
Outputs
View
Executive book review
Workflow
Tracked leadership asks
MCP
Asks queryable by agents
  • Quick
  • BigQuery
  • quick.db
  • MCP
  • GitHub
04

Concessions Ledger

A layer over warehouse deal data that captures what CRM couldn't: the gross value of commercial concessions and their cost to billable capacity.

Data overlay · Decision capture
Problem

CRM quoting couldn't capture discounts, so the gross value of concessions was invisible, and so was the tradeoff when billable people took concession work.

What I built

A filterable deal table with summary metrics and a right-side drawer. In the drawer you flag a concession, enter gross, discount, and net, and link the supporting documentation.

Impact

It made the unseeable visible. Leadership can report closed won gross and net, plus the concessions funded at the cost of other billable work.

Before: Concessions invisible in CRM
After: Gross · discount · net tracked
Architecture / Concessions LedgerInputs → Core → Outputs
Sources
SQL
Closed won (BigQuery)
Input
Concession overrides
Link
Reference docs
Override logicWarehouse + overrides
Filters & metrics
Slice deals; summarize gross, net, and concessions
Deal table
Closed won records from CRM via warehouse
Right drawer
Flag concession · gross revenue · discount · net revenue · link rationale
Overrides persisted as JSON in quick.db
Outputs
Report
Concessions over time
Insight
Capacity tradeoff visibility
MCP
Data queryable by agents
  • Quick
  • BigQuery
  • quick.db
  • MCP
  • GitHub
05

Resource Planner

The only upstream resource planning tool for Professional Services. It turns open pipeline into weekly hours of demand by craft and compares that to real capacity.

Forecasting · Capacity planning
Problem

No upstream planning existed. Craft leads were caught off guard when closed deals outstripped availability.

What I built

EMs shape open deals with craft types and hours, and craft leads enter weekly capacity. A planning view flags weekly gaps across forecasts weighted by deal-stage probability.

Impact

Gaps are visible weeks ahead and can be planned for. Craft leads aren't surprised anymore, and leadership uses the view to make headcount and efficiency investment decisions.

Before: Surprise capacity crunches
After: Weekly gaps forecast ahead
Architecture / Resource PlannerInputs → Core → Outputs
Sources
SQL
Open deals (BigQuery)
Input
EM deal shaping
Input
Craft lead capacity
Demand vs capacity modelWeekly
Deal shaping
Multiple rows per deal: craft type + hours
Forecast scenarios
Weighted by deal-stage close probability
Gap detection
Demand vs availability, flagged in hours
Filters
Regional demand, narrowed by craft
Allocations & capacity persisted as JSON in quick.db
Outputs
View
Weekly gap planner
Decision
Craft lead resourcing
Case
Headcount investment
  • Quick
  • BigQuery
  • quick.db
  • MCP
  • GitHub
06

Agentic Slide Studio

A codebase for agentic slide creation. Connect your context, prompt from any harness, and get an on-brand deck built from a fixed design library.

Codebase · Design system
Problem

On-brand customer decks meant hours of moving content around Google Slides by hand, with inconsistent results.

What I built

A GitHub repo with markdown agent instructions and design patterns. It renders slides as a Quick site with grid navigation and PDF export, and pages can be ported to Google Slides.

Impact

Idea to on-brand deck in minutes, not hours. Agentic edits are just as fast, and the federated repo keeps everyone on the latest version.

Before: Hours in Google Slides
After: One prompt → deck in minutes
Architecture / Slide StudioInputs → Core → Outputs
Sources
Input
Context & materials
Markdown
Agent instructions
Library
Design patterns
Slide generatorAny harness
Prompted authoring
Create and edit slides agentically
Web renderer
Slides as an on-brand Quick site
Grid view
Jump from full page to slide navigation
PDF export
Print the web view to PDF
Outputs
Deck
On-brand slide site
Export
Google Slides
Export
PDF
  • LLMs
  • Pi Harness
  • Quick
  • GitHub
  • Markdown
  • Frontend code
  • Any harness

And one more: this site.

I built this portfolio agentically with Vercel v0. It started as a written document and became a published site, so the site is its own case study.

07

This Portfolio, Built with v0

The site you are reading is itself an agentic build. I gave Vercel v0 a written description of my systems, and it turned that into the architecture-first portfolio you see here.

Meta example · Vercel v0
Problem

Six systems, described in a plain written document, needed to become a visual, engaging story about technical depth and business impact. No designer or front-end sprint was involved.

What I built

I directed v0 to parse the document, structure each system into problem, architecture, and impact, and design a diagram-led site with live SVG wiring between nodes. I then iterated through prompts.

Impact

I went from a document to a published site in a single conversation. It applies the same build loop as every other system on this page: find the toil, build agentically, and ship.

Before: A written doc in a folder
After: A live, interactive portfolio
Architecture / This SiteInputs → Core → Outputs
Sources
Doc
Written system notes (.docx)
Prompt
Builder direction & taste
Feedback
Iterative refinement
v0 agentVercel v0
Document parsing
Extracted six systems into structured data
Design direction
Architecture-diagram visual language
Next.js build
React components with live SVG connectors
Browser verification
Rendered, inspected, and refined in a sandbox
Outputs
Site
This portfolio
Deploy
Published on Vercel
Source
Git-backed codebase
  • Vercel v0
  • Next.js
  • React
  • TypeScript
  • Vercel

The build loop behind every system.

A repeatable pattern: find the toil, build it agentically, ship it, and let the community make it better.

  1. 01

    Find the toil

    Spot work where hours go into production instead of judgment.

  2. 02

    Wire the context

    Connect MCP tools, the data warehouse, and domain corpora.

  3. 03

    Build agentically

    Prompt, generate, and iterate across harnesses, models, and v0.

  4. 04

    Ship always-on

    Host on Quick with real-time SQL and persistent state.

  5. 05

    Federate on GitHub

    Anyone can consume, extend, and improve the system through PRs.